Why is DeepSeek so good? @SerranoAcademy
Why is DeepSeek so good?  @SerranoAcademy
Uploaded February 2025 | Updated November 2026, 3 hours ago
DeepSeek has performed much better than ChatGPT and with a much smaller model. One of the reasons is: reinforcement learning!
Why is DeepSeek so good?The Discrete Fourier TransformNewtons method for approximating zeros of polynomials - Math for ML with Deeplearning.aiProximal Policy Optimization (PPO) - How to train Large Language ModelsMath and OCD - My story with the Thue-Morse sequenceThank you for 100K subscribers! I’m planning tons of new content coming soon, so excited!A friendly introduction to Recurrent Neural NetworksThe math behind Attention: Keys, Queries, and Values matricesWill AI help us, or make us dependent? - A Tale of Two CitiesThe covariance matrixGRPO - Group Relative Policy Optimization  - How DeepSeek trains reasoning modelsReinforcement Learning with Human Feedback (RLHF) - How to train and fine-tune Transformer Models
Luis Serrano Academy |

Why is DeepSeek so good?

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER